Mailing-List: contact dev-help@harmony.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@harmony.apache.org
Received-SPF: pass (nike.apache.org: domain of aleksey.shipilev@gmail.com
 designates 209.85.198.238 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=message-id:date:from:to:subject:in-reply-to:mime-version
         :content-type:content-transfer-encoding:content-disposition
         :references;
        b=nKNPs4+iNxYilykT2rwD4FI04uMkwClkL5fpK5ngKcxUclkfyuB65OV0nFeszySPTR
         okcM6Eyha5s4BnyeD+mDkcex6fNRATcTFVl0wmu+pXgEYRysuVlSzAsQM8naynYdtu1a
         qMCYO0lIul27BZd1YsmNoCtM/0mg4zVFn7X48=
Message-ID: <4bebff790901071450k238f4b82rfb3ab6bf8cdeb55e@mail.gmail.com>
Date: Thu, 8 Jan 2009 01:50:21 +0300
From: "Aleksey Shipilev" <aleksey.shipilev@gmail.com>
To: dev@harmony.apache.org
Subject: Re: [jira] Commented: (HARMONY-6056)
 [classlib][jit][opt][performance] Optimize heap allocations in String(String,
 int) constructor for JIT
In-Reply-To: <4964C7C7.5060005@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
References: <1100194590.1229915264310.JavaMail.jira@brutus>
	 <49092120.1231268324402.JavaMail.jira@brutus>
	 <70c713190901061950l53b326c0hc19b4adaf2469877@mail.gmail.com>
	 <4bebff790901070102w4183c200v16af97378e7843cd@mail.gmail.com>
	 <4964C05B.2080801@gmail.com>
	 <4bebff790901070655v28fb59c0od8884af26e110e78@mail.gmail.com>
	 <4964C7C7.5060005@gmail.com>

Hi,

I had some spare time and made the benchmark [1] for this issue (never
do the performance optimizations without the benchmarks covering your
case :)).

Presumably the good patch for the issue should increase the score on
this benchmark. If my understanding is good, the boost should be
observed on J9 with Kevin's version of the patch. The boost on DRLVM
should be observed with the final patch.

Thanks,
Aleksey.

On Harmony-M7 at P-D 2.8 Ghz / PC2-5300 / Gentoo Linux x86:

$ harmony-M7/bin/ava -Xms1024M -Xmx1024M -server -jar StringBench.jar
(String)base + (int)add:
-------------------------------------------
 base length (vars with rows): 0..2..10
  add length (vars with cols): 0..2..10

 loop duration = 100 msecs
 target variance = 0.05

ops/msec, the more the better:
4369, 2842, 1976, 1563, 1226,
4462, 2829, 1966, 1555, 1225,
4430, 2836, 1956, 1563, 1223,
4394, 2831, 1956, 1558, 1217,
4357, 2821, 1954, 1551, 1217,

Thanks,
Aleksey.

[1] https://issues.apache.org/jira/secure/attachment/12397340/StringBench.jar

On Wed, Jan 7, 2009 at 6:18 PM, Tim Ellison <t.p.ellison@gmail.com> wrote:
> Aleksey Shipilev wrote:
>> Ok, we can implement the in-place Integer.toString() and specialize
>> the radix-10 conversion in Integer. Then Classlib performance guys
>> might use the inplace conversion to optimize StringBuilder performance
>> or even catch the concatenation like J9 does.
>>
>> My idea is to share whatever optimization between all VMs that use the Classlib.
>
> indeed!
>
>